Autophagy is important in cancer development, progression and response to therapy. Although we are already trying to target autophagy in the clinic, there is currently no way to identify the tumors that will or will not benefit from autophay inhibition. We recently reported that some breast tumor cells are much more dependent on autophagy than others and our data suggest that it is only in the truly autophagy-dependent tumor cells (which we define as those that require autophagy to survive even in the absence of additional stress) where autophagy inhibition will be effective on its own or in combination with other agents. Indeed, our data suggest that combining autophagy inhibition with another agent in tumor cells that are not autophagy-dependent can be counterproductive. We also identified a potential mechanism to explain autophagy-dependent breast cancer- autophagy controls STAT3 activity in only some tumor cells and this is necessary and sufficient to explain autophagy-dependent survival; moreover recent studies have also indicated that tumor stem-like cell activity requires autophagy in these cells. These data led us to hypothesize: autophagy-dependency defines a specific tumor subtype whose cell survival and stem cell-like characteristics depend on autophagy: these are the tumors that will respond best to autophagy inhibition. To test our hypothesis, we have three specific aims. Aim 1. Identify autophagy-dependent tumors and determine response to autophagy inhibition. We hypothesize that we can identify autophagy-dependent tumors and will test a preliminary gene expression signature that we think can do this. We propose that autophagy inhibition will be effective but only for autophagy-dependent tumors and will test this in primary tumors and by neoadjuvant and adjuvant treatment of surgically resected metastatic tumors. Aim 2. Test if autophagy inhibition sensitizes to other treatments in autophagy-dependent breast cancer. To test the hypothesis that autophagy-dependent tumors will benefit most from combination treatment with autophagy inhibitors, experiments using large scale shRNA analysis well as a novel bioinformatics approach using the COXEN principle will work out how best to target autophagy in combination with other drugs. Aim 3. Determine the mechanism underlying autophagy-dependency in breast tumors. We hypothesize that autophagy regulation of autocrine signaling explains STAT3 activation, survival and stem cell like of autophagy-dependent tumor cells. We will test this by analyzing autophagy's ability to signal through cytokines like IL6, which we have shown is specifically controlled by autophagy only in the autophagy-dependent tumor cells. If our ideas are correct, our work will help define a novel subtype of autophagy-dependent breast cancer and provide a new way to identify, treat and improve combination treatments for these tumors while providing new insights into the roles of autophagy in cancer therapy and cancer biology.